TRICHOTOMY, STABILITY, AND OSCILLATION OF
A FUZZY DIFFERENCE EQUATION
G. STEFANIDOU AND G. PAPASCHINOPOULOS
Received 10 November 2003
We study the trichotomy character, the stability, and the oscillatory behavior of the positive solutions of a fuzzy difference equation.
1. Introduction
Difference equations have already been successfully applied in a number of sciences (for
a detailed study of the theory of difference equations and their applications, see [1, 2, 7,
8, 11].
The problem of identifying, modeling, and solving a nonlinear difference equation
concerning a real-world phenomenon from experimental input-output data, which is
uncertain, incomplete, imprecise, or vague, has been attracting increasing attention in
recent years. In addition, nowadays, there is an increasing recognition that for understanding vagueness, a fuzzy approach is required. The effect is the introdution and the
study of the fuzzy difference equations (see [3, 4, 13, 14, 15]).
In this paper, we study the trichotomy character, the stability, and the oscillatory behavior of the positive solutions of the fuzzy difference equation
k
ci xn− pi
,
d
j =1 j xn−q j
xn+1 = A + mi=1
(1.1)
where k,m ∈ {1,2,... }, A,ci ,d j , i ∈ {1,2,...,k}, j ∈ {1,2,...,m}, are positive fuzzy numbers, pi , i ∈ {1,2,...,k}, q j , j ∈ {1,2,...,m}, are positive integers such that p1 < p2 <
· · · < pk , q1 < q2 < · · · < qm , and the initial values xi , i ∈ {−π, −π + 1,...,0}, where
π = max pk , qm ,
(1.2)
are positive fuzzy numbers.
Studying a fuzzy difference equation results concerning the behavior of a related family
of systems of parametric ordinary difference equations is required. Some necessary results
Copyright © 2004 Hindawi Publishing Corporation
Advances in Difference Equations 2004:4 (2004) 337–357
2000 Mathematics Subject Classification: 39A10
URL: http://dx.doi.org/10.1155/S1687183904311015
338
Fuzzy difference equations
concerning the corresponding family of systems of ordinary difference equations of (1.1)
have been proved in [16] and others are given in this paper.
2. Preliminaries
We need the following definitions.
For a set B, we denote by B the closure of B. We say that a fuzzy set A, from R+ = (0, ∞)
into the interval [0,1], is a fuzzy number, if A is normal, convex, upper semicontinu
ous (see [14]), and the support suppA = a∈(0,1] [A]a = {x : A(x) > 0} is compact. Then
from [12, Theorems 3.1.5 and 3.1.8], the a-cuts of the fuzzy number A, [A]a = {x ∈ R+ :
A(x) ≥ a}, are closed intervals.
We say that a fuzzy number A is positive if suppA ⊂ (0, ∞).
It is obvious that if A is a positive real number, then A is a positive fuzzy number and
[A]a = [A,A], a ∈ (0,1]. In this case, we say that A is a trivial fuzzy number.
We say that xn is a positive solution of (1.1) if xn is a sequence of positive fuzzy numbers
which satisfies (1.1).
A positive fuzzy number x is a positive equilibrium for (1.1) if
k
ci x
.
d
j =1 j x
x = A + mi=1
(2.1)
Let E, H be fuzzy numbers with
[E]a = [El,a ,Er,a ],
[H]a = [Hl,a ,Hr,a ],
a ∈ (0,1].
(2.2)
According to [10] and [13, Lemma 2.3], we have that MIN{E,H } = E if
El,a ≤ Hl,a ,
Er,a ≤ Hr,a ,
a ∈ (0,1].
(2.3)
Moreover, let ci , fi ,d j ,g j , i = 1,2,...,k, j = 1,2,...,m, be positive fuzzy numbers such
that for a ∈ (0,1],
= ci,l,a ,ci,r,a ,
d j a = d j,l,a ,d j,r,a ,
k
ci
E = mi=1 ,
ci
a
j =1 d j
= fi,l,a , fi,r,a ,
g j a = g j,l,a ,g j,r,a ,
k
fi
H = mi=1 .
fi
a
j =1 g j
(2.4)
(2.5)
We will say that E is less than H and we will write
E≺H
(2.6)
if
k
i=1 supa∈(0,1] ci,r,a
m
j =1 inf a∈(0,1] d j,l,a
k
inf a∈(0,1] fi,l,a
< mi=1
.
j =1 supa∈(0,1] g j,r,a
(2.7)
G. Stefanidou and G. Papaschinopoulos 339
In addition, we will say that E is equal to H and we will write
.
E=H
if E ≺ H, H ≺ E,
(2.8)
which means that for i = 1,2,...,k, j = 1,2,...,m, and a ∈ (0,1],
ci,l,a = ci,r,a ,
fi,l,a = fi,r,a ,
d j,l,a = d j,r,a ,
g j,l,a = g j,r,a ,
(2.9)
and so
El,a = Er,a = Hl,a = Hr,a ,
a ∈ (0,1],
(2.10)
which imples that E, H are equal real numbers.
For the fuzzy numbers E, H, we give the metric (see [9, 17, 18])
D(E,H) = supmax El,a − Hl,a , Er,a − Hr,a ,
(2.11)
where sup is taken for all a ∈ (0,1].
The fuzzy analog of boundedness and persistence (see [5, 6]) is given as follows: we
say that a sequence of positive fuzzy numbers xn persists (resp., is bounded) if there exists
a positive number M (resp., N) such that
supp xn ⊂ [M, ∞)
resp., suppxn ⊂ (0,N] ,
n = 1,2,....
(2.12)
In addition, we say that xn is bounded and persists if there exist numbers M,N ∈ (0, ∞)
such that
suppxn ⊂ [M,N],
n = 1,2,....
(2.13)
Let xn be a sequence of positive fuzzy numbers such that
xn
a
= Ln,a ,Rn,a ,
a ∈ (0,1], n = 0,1,...,
(2.14)
and let x be a positive fuzzy number such that
[x]a = [La ,Ra ],
a ∈ (0,1].
(2.15)
We say that xn nearly converges to x with respect to D as n → ∞ if for every δ > 0, there
exists a measurable set T, T ⊂ (0,1], of measure less than δ such that
limDT xn ,x = 0,
as n −→ ∞,
(2.16)
where
DT xn ,x =
sup
a∈(0,1]−T
max Ln,a − La , Rn,a − Ra .
If T = ∅, we say that xn converges to x with respect to D as n → ∞.
(2.17)
340
Fuzzy difference equations
Let X be the set of positive fuzzy numbers. Let E,H ∈ X. From [18, Theorem 2.1], we
have that El,a , Hl,a (resp., Er,a , Hr,a ) are increasing (resp., decreasing) functions on (0,1].
Therefore, using the definition of the fuzzy numbers, there exist the Lebesque integrals
J
El,a − Hl,a da,
J
Er,a − Hr,a da,
(2.18)
where J = (0,1]. We define the function D1 : X × X → R+ such that
D1 (E,H) = max
El,a − Hl,a da,
J
J
Er,a − Hr,a da .
(2.19)
If D1 (E,H) = 0, we have that there exists a measurable set T of measure zero such that
El,a = Hl,a ,
Er,a = Hr,a
∀a ∈ (0,1] − T.
(2.20)
We consider, however, two fuzzy numbers E, H to be equivalent if there exists a measurable set T of measure zero such that (2.20) hold and if we do not distinguish between
equivalence of fuzzy numbers, then X becomes a metric space with metric D1 .
We say that a sequence of positive fuzzy numbers xn converges to a positive fuzzy
number x with respect to D1 as n → ∞ if
limD1 xn ,x = 0,
as n −→ ∞.
(2.21)
We define the fuzzy analog for periodicity (see [11]) as follows.
A sequence {xn } of positive fuzzy numbers xn is said to be periodic of period p if
D xn+p ,xn = 0,
n = 0,1,....
(2.22)
Suppose that (1.1) has a unique positive equilibrium x. We say that the positive equithere exists a δ = δ() such that for everyposlibrium x of (1.1) is stable if for every > 0,
itive solution xn of (1.1) which satisfies D x−i ,x ≤ δ, i = 0,1,...,π, we have D xn ,x ≤ for all n ≥ 0.
Moreover, we say that the positive equilibrium x of (1.1) is nearly asymptotically stable
if it is stable and every positive solution of (1.1) nearly tends to the positive equilibrium
of (1.1) with respect to D as n → ∞.
Finally, we give the fuzzy analog of the concept of oscillation (see [11]). Let xn be a
sequence of positive fuzzy numbers and let x be a positive fuzzy number. We say that xn
oscillates about x if for every n0 ∈ N, there exist s,m ∈ N, s,m ≥ n0 , such that
MIN xm ,x = xm ,
MIN xs ,x = x
(2.23)
or
MIN xm ,x = x,
MIN xs ,x = xs .
(2.24)
G. Stefanidou and G. Papaschinopoulos 341
3. Main results
Arguing as in [13, 14, 15], we can easily prove the following proposition which concerns
the existence and the uniqueness of the positive solutions of (1.1).
Proposition 3.1. Consider (1.1), where k,m ∈ {1,2,... }, A,ci ,d j , i ∈ {1,2,...,k}, j ∈ {1,
2,...,m}, are positive fuzzy numbers, and pi , q j , i ∈ {1,2,...,k}, j ∈ {1,2,...,m}, are positive integers. Then for any positive fuzzy numbers x−π ,x−π+1 ,...,x0 , there exists a unique
positive solution xn of (1.1) with initial values x−π ,x−π+1 ,...,x0 .
Now, we present conditions so that (1.1) has unbounded solutions.
Proposition 3.2. Consider (1.1), where k,m ∈ {1,2,... }, A,ci ,d j , i ∈ {1,2,...,k}, j ∈ {1,
2,...,m}, are positive fuzzy numbers, and pi , i ∈ {1,2,...,k}, q j , j ∈ {1,2,...,m}, are positive integers. If
k
ci
G = mi=1 ,
A ≺ G,
(3.1)
j =1 d j
then (1.1) has unbounded solutions.
Proof. Let
[A]a = Al,a ,Ar,a ,
a ∈ (0,1].
(3.2)
From (2.4) and (3.2) and since A, ci , d j , i = 1,2,...,k, j = 1,2,...,m, are positive fuzzy
numbers, there exist positive real numbers B, C, ai , ei , h j , b j , i = 1,2,...,k, j = 1,2,...,m,
such that
B = inf Al,a ,
C = sup Ar,a ,
ei = sup ci,r,a ,
h j = inf d j,l,a ,
a∈(0,1]
a∈(0,1]
a∈(0,1]
a∈(0,1]
ai = inf ci,l,a ,
a∈(0,1]
b j = sup d j,r,a .
(3.3)
a∈(0,1]
Let xn be a positive solution of (1.1) such that (2.14) hold and the initial values xi , i =
−π, −π + 1,...,0, are positive fuzzy numbers which satisfy
xi
a
= Li,a ,Ri,a ,
i = −π, −π + 1,...,0, a ∈ (0,1]
(3.4)
and for a fixed ā ∈ (0,1], the relations
Ri,ā >
Z2
,
W −C
Li,ā < W,
i = −π, −π + 1,...,0,
(3.5)
are satisfied, where
k
ei
Z = mi=1 ,
j =1 h j
k
ai
W = mi=1 .
j =1 b j
(3.6)
342
Fuzzy difference equations
Using [15, Lemma 1], we can easily prove that Ln,a , Rn,a satisfy the family of systems of
parametric ordinary difference equations
k
ci,l,a Ln− pi ,a
,
Ln+1,a = Al,a + mi=1
d
j =1 j,r,a Rn−q j ,a
k
Rn+1,a = Ar,a +
ci,r,a Rn− pi ,a
mi=1
,
j =1 d j,l,a Ln−q j ,a
n = 0,1,....
(3.7)
Since (3.1) holds, it is obvious that
k
ci,r,ā
.
d
j =1 j,l,ā
Al,ā < mi=1
(3.8)
Using (3.8) and applying [16, Proposition 1] to the system (3.7) for a = ā, we have that
lim Ln,ā=Al,ā ,
n→∞
lim Rn,ā = ∞.
(3.9)
n→∞
Therefore, from (3.9), the solution xn of (1.1) which satisfies (3.4) and (3.5) is un
bounded.
Remark 3.3. From the proof of Proposition 3.2, it is obvious that (1.1) has unbounded
solutions if there exists at least one a ∈ (0,1] such that (3.8) holds.
In the following proposition, we study the boundedness and persistence of the positive
solutions of (1.1).
Proposition 3.4. Consider (1.1), where k,m ∈ {1,2,... }, A,ci ,d j , i ∈ {1,2,...,k}, j ∈
{1,2,...,m}, are positive fuzzy numbers, and pi , i ∈ {1,2,...,k }, q j , j ∈ {1,2,...,m}, are
positive integers. If either
.
A=G
(3.10)
G≺A
(3.11)
or
holds, then every positive solution of (1.1) is bounded and persists.
Proof. Firstly, suppose that (3.10) is satisfied; then A, ci , d j , i = 1,2,...,k, j = 1,2,...,m,
are positive real numbers. Hence, for i = 1,2,...,k, j = 1,2,...,m, we get
A = Al,a = Ar,a ,
ci = ci,l,a = ci,r,a ,
d j = d j,l,a = d j,r,a ,
k
ci
A = mi=1 .
j =1 d j
a ∈ (0,1],
(3.12)
(3.13)
G. Stefanidou and G. Papaschinopoulos 343
Let xn be a positive solution of (1.1) such that (2.14) hold and let xi , i = −π, −π +
1,...,0, be the positive initial values of xn such that (3.4) hold. Then there exist positive
numbers Ti , Si , i = −π, −π + 1,...,0, such that
Ti ≤ Li,a ,Ri,a ≤ Si ,
i = −π, −π + 1,...,0.
(3.14)
Let (yn ,zn ) be the positive solution of the system of ordinary difference equations
k
ci yn− pi
,
d
j =1 j zn−q j
yn+1 = A + mi=1
k
ci zn− pi
,
d
j =1 j y n − q j
zn+1 = A + mi=1
(3.15)
with initial values (yi ,zi ), i = −π, −π + 1,...,0, such that yi = Ti , zi = Si , i = −π, −π +
1,...,0. Then from (3.14) and (3.15), we can easily prove that
y1 ≤ L1,a ,
R1,a ≤ z1 ,
a ∈ (0,1],
(3.16)
n = 1,2,..., a ∈ (0,1].
(3.17)
and working inductively, we take
yn ≤ Ln,a ,
Rn,a ≤ zn ,
Since from (3.13) and [16, Proposition 3], (yn , zn ) is bounded and persists, from (3.17),
it is obvious that xn is also bounded and persists.
Now, suppose that (3.11) holds; then
B > Z,
C > W.
(3.18)
We concider the system of ordinary difference equations
k
ai y n − p i
,
j =1 b j zn−q j
yn+1 = B + mi=1
k
ei zn− pi
,
j =1 h j y n − q j
zn+1 = C + mi=1
(3.19)
where B,C,ai ,ei ,b j ,h j , i = 1,2,...,k, j = 1,2,...,m, are defined in (3.3).
Let (yn ,zn ) be a solution of (3.19) with initial values (yi ,zi ), i = −π, −π + 1,...,0, such
that yi = Ti , zi = Si , i = −π, −π + 1,...,0, where Ti ,Si , i = −π, −π + 1,...,0, are defined
in (3.14). Arguing as above, we can prove that (3.17) holds. Since from (3.18) and [16,
Proposition 3], (yn , zn ) is bounded and persists, then from (3.17), it is obvious that, xn is
also bounded and persists. This completes the proof of the proposition.
In what follows, we need the following lemmas.
Lemma 3.5. Let ri ,s j , i = 1,2,...,k, j = 1,2,...,m, be positive integers such that
r1 ,r2 ,...,rk ,s1 ,s2 ,...,sm = 1,
(3.20)
where (r1 ,r2 ,...,rk ,s1 ,s2 ,...,sm ) is the greatest common divisor of the integers ri ,s j , i = 1,2,
...,k, j = 1,2,...,m. Then the following statements are true.
344
Fuzzy difference equations
(I) There exists an even positive integer w1 such that for any nonnegative integer p, there
exist nonnegative integers αip , β j p , i = 1,2,...,k, j = 1,2,...,m, such that
k
αip ri +
i =1
m
β j p s j = w1 + 2p,
p = 0,1,...,
(3.21)
j =1
where mj=1 β j p is an even integer.
(II) Suppose that all ri , i = 1,2,...,k, are not even and all s j , j = 1,2,...,m, are not odd
integers. Then there exists an odd positive integer w2 such that for any nonnegative integer p,
there exist nonnegative integers γip , δ j p , i = 1,2,...,k, j = 1,2,...,m, such that
k
γip ri +
i =1
m
δ j p s j = w2 + 2p,
p = 0,1,...,
(3.22)
j =1
where mj=1 δ j p is an even integer.
(III) Suppose that all ri , i = 1,2,...,k, are not even and all s j , j = 1,2,...,m, are not odd
integers. Then there exists an even positive integer w3 such that for any nonnegative integer p,
there exist nonnegative integers ip , ξ j p , i = 1,2,...,k, j = 1,2,...,m, such that
k
ip ri +
i =1
m
ξ j p s j = w3 + 2p,
p = 0,1,...,
(3.23)
j =1
where mj=1 ξ j p is an odd integer.
(IV) There exists an odd positive integer w4 such that for any nonnegative integer p, there
exist nonnegative integers λip , µ j p , i = 1,2,...,k, j = 1,2,...,m, such that
k
λip ri +
i =1
where
m
j =1 µ j p
m
µ j p s j = w4 + 2p,
p = 0,1,...,
(3.24)
j =1
is an odd integer.
Proof. (I) Since(3.20) holds, there exist integers ηi , ι j , i = 1,2,...,k, j = 1,2,...,m, such
that
k
ηi ri +
i =1
m
ι j s j = 1.
(3.25)
j =1
If for any real number a, we denote by [a] the integral part of a, we set for i = 2,3,...,k,
j = 1,2,...,m,
α1p = 2pη1 + 2
k
ri + 2
i =2
αip = 2pηi + 2gip r1 ,
m
j =1
sj − 2
k
i=2
gip ri − 2
m
j =1
β j p = 2pι j + 2h j p r1 ,
h j ps j ,
(3.26)
G. Stefanidou and G. Papaschinopoulos 345
where
gip =
− pηi
r1
+ 1,
hjp =
− pι j
i = 2,3,...,k, j = 1,2,...,m.
+ 1,
r1
(3.27)
Therefore, from (3.25) and (3.26), we can easily prove that αip ,β j p , i = 1,2,...,k, j =
1,2,...,m, which are defined in (3.26), are positive integers satisfying (3.21) for
w1 = 2r1
k
ri +
i =2
m
sj
(3.28)
j =1
and mj=1 β j p is an even number.
(II) Firstly, suppose that one of ri , i = 1,2,...,k, is an odd positive integer and without
loss of generality, let r1 be an odd positive integer. Relation (3.22) follows immediately if
we set for i = 2,...,k and for j = 1,2,...,m,
γ1p = α1p + 1,
γip = αip ,
δ j p = β j p,
w2 = w1 + r1 .
(3.29)
Now, suppose that ri , i = 1,2,...,k, are even positive integers; then from (3.20), one
of s j , j = 1,2,...,m, is an odd positive integer and from the hypothesis, one of s j , j =
1,2,...,m, is an even positive integer. Without loss of generality, let s1 be an odd positive
integer and s2 be an even positive integer. Relation (3.22) follows immediately if we set
for i = 1,2,...,k and for j = 3,...,m,
γip = αip ,
δ1p = β1p + 1,
δ2p = β2p + 1,
δ j p = β j p,
w2 = w1 + s1 + s2 .
(3.30)
(III) Firstly, suppose that one of s j , j = 1,2,...,m, is an even positive integer and without loss of generality, let s1 be an even positive integer. Relation (3.23) follows immediately if we set for i = 1,2,...,k and j = 2,...,m,
ip = αip ,
ξ1p = β1p + 1,
ξ j p = β j p,
w3 = w1 + s1 .
(3.31)
Now, suppose that s j , j = 1,2,...,m, are odd positive integers; then from the hypothesis, at least one of ri , i = 1,2,...,k, is an odd positive integer, and without loss of generality, let r1 be an odd integer. Relation (3.23) follows immediately if we set for i = 2,...,k,
j = 2,3,...,m,
1p = α1p + 1,
ip = αip ,
δ1p = β1p + 1,
δ j p = β j p,
w3 = w1 + s1 + r1 .
(3.32)
(IV) Firstly, suppose that at least one of s j , j = 1,2,...,m, is an odd positive integer
and without loss of generality, let s1 be an odd positive integer. Relation (3.24) follows
immediately if we set for i = 1,2,...,k, j = 2,3,...,m,
λip = αip ,
µ1p = β1p + 1,
µ j p = β j p,
w4 = w1 + s1 .
(3.33)
346
Fuzzy difference equations
Now, suppose that s j , j = 1,2,...,m, are even positive integers; then from (3.20), at
least one of ri , i = 1,2,...,k, is an odd positive integer, and without loss of generality, let r1
be an odd positive integer. Relation (3.24) follows immediately if we set for i = 2,3,...,k,
j = 2,3,...,m,
λ1p = α1p + 1,
λip = αip ,
µ1p = β1p + r1 ,
µ j p = β j p,
w4 = w1 + r1 s1 + 1 .
(3.34)
This completes the proof of the lemma.
Lemma 3.6. Consider system (3.19), where B,C are positive constants such that
k
ei
B = mi=1 ,
j =1 h j
k
ai
C = mi=1 .
(3.35)
j =1 b j
Then the following statements are true.
(I) Let r be a common divisor of the integers pi + 1, q j + 1, i = 1,2,...,k, j = 1,2,...,m,
such that
pi + 1 = rri ,
i = 1,2,...,k,
q j + 1 = rs j ,
j = 1,2,...,m;
(3.36)
then system (3.19) has periodic solutions of prime period r. Moreover, if all ri , i = 1,2,...,k,
(resp., s j , j = 1,2,...,m) are even (resp., odd) positive integers, then system (3.19) has periodic solutions of prime period 2r.
(II) Let r be the greatest common divisor of the integers pi + 1, q j + 1, i = 1,2,...,k, j =
1,2,...,m, such that (3.36) hold; then if all ri , i = 1,2,...,k, (resp., s j , j = 1,2,...,m) are even
(resp., odd) positive integers, every positive solution of (3.19) tends to a periodic solution of
period 2r; otherwise, every positive solution of (3.19) tends to a periodic solution of period r.
Proof. (I) From relations (3.35), (3.36), and [16, Proposition 2], system (3.19) has periodic solutions of prime period r.
Now, we prove that system (3.19) has periodic solutions of prime period 2r, if all ri ,
i = 1,2,...,k, (resp., s j , j = 1,2,...,m) are even (resp., odd) positive integers.
Suppose first that pk < qm . Let (yn ,zn ) be a positive solution of (3.19) with initial values
satisfying
y−rsm +2rλ+ζ = y−r+ζ ,
z−rsm +2rλ+ζ = z−r+ζ ,
y−rsm +2rν+r+ζ = y−2r+ζ ,
z−rsm +2rν+r+ζ = z−2r+ζ ,
s −1
s −3
λ = 0,1,..., m
,
ν = 0,1,..., m
,
ζ = 1,2,...,r,
2
2
(3.37)
and, in addition, for ζ = 1,2,...,r,
y−2r+ζ > B,
y−r+ζ > B,
z−r+ζ =
C y−2r+ζ
,
y−2r+ζ − B
z−2r+ζ =
C y−r+ζ
.
y−r+ζ − B
(3.38)
G. Stefanidou and G. Papaschinopoulos 347
From (3.19), (3.35), (3.36), (3.37), and (3.38), we get for ζ = 1,2,...,r,
y−2r+ζ
= y−2r+ζ ,
z−r+ζ
y−r+ζ
= B+C
= y−r+ζ ,
z−2r+ζ
z−2r+ζ
= z−2r+ζ ,
y−r+ζ
z−r+ζ
= C+B
= z−r+ζ .
y−2r+ζ
yζ = B + C
zζ = C + B
yr+ζ
zr+ζ
(3.39)
Let a v ∈ 2,3,... . Suppose that for all u = 1,2,...,v − 1 and ζ = 1,2,...,r, we have
y2ur+ζ = y−2r+ζ ,
z2ur+ζ = z−2r+ζ ,
y2ur+r+ζ = y−r+ζ ,
z2ur+r+ζ = z−r+ζ . (3.40)
Then from (3.19), (3.35)–(3.40), we get for ζ = 1,2,...,r,
y2vr+ζ = B + C
y−2r+ζ
= y−2r+ζ .
z−r+ζ
(3.41)
Similarly, we can prove that for ζ = 1,2,...,r,
z2vr+ζ = z−2r+ζ ,
y2vr+r+ζ = y−r+ζ ,
z2vr+r+ζ = z−r+ζ .
(3.42)
Therefore, from (3.39)–(3.42), we have that system (3.19) has periodic solutions of
period 2r.
Now, suppose that qm < pk . Let (yn ,zn ) be a positive solution of (3.19) such that the
initial values satisfy relations (3.38) and for ω = 0,1,...,rk /2 − 1, θ = 1,2,...,2r,
y−rrk +2rω+θ = y−2r+θ ,
z−rrk +2rω+θ = z−2r+θ .
(3.43)
Then arguing as above, we can easily prove that (yn ,zn ) is a periodic solution of period 2r.
This completes the proof of statement (I).
(II) Now, we prove that every positive solution of system (3.19) tends to a periodic
solution of period κr, where

2
if ri , i = 1,2,...,k, are even, s j , j = 1,2,...,m, are odd,
κ=
1 otherwise.
(3.44)
Let (yn ,zn ) be an arbitrary positive solution of (3.19). We prove that there exist the
lim yκnr+i = i ,
n→∞
i = 0,1,...,κr − 1.
(3.45)
We fix a τ ∈ {0,1,...,κr − 1}. Since from [16, Proposition 3], the solution (yn ,zn ) is
bounded and persists, we have
liminf yκnr+τ = lτ ≥ B,
n→∞
limsup yκnr+τ = Lτ < ∞,
n→∞
liminf zκnr+τ = mτ ≥ C,
n→∞
limsup zκnr+τ = Mτ < ∞.
(3.46)
n→∞
Therefore, from relations (3.19), (3.35), and (3.46), we take
mτ =
CLτ
,
Lτ − B
lτ =
BMτ
.
Mτ − C
(3.47)
348
Fuzzy difference equations
We prove that (3.45) is true for i = τ. Suppose on the contrary that lτ < Lτ . Then from
(3.46), there exists an > 0 such that
Lτ > lτ + > B + .
(3.48)
In view of (3.46), there exists a sequence nµ , µ = 1,2,..., such that
lim yκrnµ +τ = Lτ ,
lim yr(κnµ −ri )+τ = Tri ,τ ≤ Lτ ,
µ→∞
µ→∞
(3.49)
lim zr(κnµ −s j )+τ = Ss j ,τ ≥ mτ .
µ→∞
In view of (3.19), (3.35), (3.46), (3.47), and (3.49), we take
k
ai Tri ,τ
CLτ
≤B+
= Lτ
Lτ = B + mi=1
j =1 b j Ss j ,τ
(3.50)
mτ
and obviously, we have that
Tri ,τ = Lτ ,
i = 1,2,...,k,
Ss j ,τ = mτ ,
(3.51)
j = 1,2,...,m.
In addition, using (3.19), (3.35), (3.46), (3.47), and (3.51), for κ = 2, from statements
(I) and (IV) of Lemma 3.5 and arguing as above, we take for γ = 0,1,...,
lim yr(2nµ −w1 −2γ)+τ = Lτ ,
lim zr(2nµ −w1 −s1 −2γ)+τ = mτ ,
µ→∞
µ→∞
(3.52)
and for κ = 1 and from all the statements of Lemma 3.5,
lim yr(nµ −w1 −2γ)+τ = Lτ ,
µ→∞
lim zr(nµ −w3 −2γ)+τ = mτ ,
µ→∞
lim yr(nµ −w2 −2γ)+τ = Lτ ,
µ→∞
lim zr(nµ −w4 −2γ)+τ = mτ ,
(3.53)
µ→∞
w1 ,w2 ,w3 ,w4 are defined in Lemma 3.5. Let a σκ ∈ {0,1,...,(3 − κ)φ}, φ = max rk ,sm . Suppose first that κ = 2. Then in view
of (3.19), there exist positive integers p, q and a continuous function Fσ2 : R × R× · · · ×
R → R such that
yr(2nµ +2σ2 )+τ = B + Fσ2 ζnµ ,0 ,... ,ζnµ ,p ,ξnµ ,0 ,...,ξnµ ,q ,
(3.54)
where for i = 0,1,..., p, j = 0,1,..., q,
ζnµ ,i = yr(2nµ −w1 −2i)+τ ,
ξnµ , j = zr(2nµ −w1 −s1 −2 j)+τ .
(3.55)
If κ = 1, there exist positive integers v1 , v2 , v3 , v4 and a continuous function Gσ1 : R ×
R× · · · × R → R such that
yr(nµ +σ1 )+τ = B + Gσ1 ζnµ ,0 ,...,ζnµ ,v1 , ζ¯nµ ,0 ,..., ζ¯nµ ,v2 ,ξnµ ,0 ,...,ξnµ ,v3 , ξ¯nµ ,0 ,..., ξ¯nµ ,v4 ,
(3.56)
G. Stefanidou and G. Papaschinopoulos 349
where for i = 0,1,...,v1 , i¯ = 0,1,...,v2 , j = 0,1,...,v3 , and j¯ = 0,1,...,v4 ,
ζ¯nµ ,i¯ = yr(nµ −w2 −2i)+τ
¯ ,
ξ¯n , j¯ = zr(n −w −2 j)+τ
¯ .
ζnµ ,i = yr(nµ −w1 −2i)+τ ,
ξnµ , j = zr(nµ −w3 −2 j)+τ ,
µ
µ
(3.57)
4
Therefore, from (3.47), (3.52), (3.53), (3.54), and (3.56), it follows that
lim yr(κnµ +κσκ )+τ = B +
µ→∞
CLτ
= Lτ .
mτ
(3.58)
Using the same argument to prove (3.58) and using (3.19), we can easily prove that for
i = 1,2,...,k, j = 1,2,...,m,
lim yr(κnµ +κσκ −ri )+τ = Lτ ,
lim zr(κnµ +κσκ −s j )+τ = mτ .
µ→∞
µ→∞
(3.59)
Therefore, if δ = (mτ − C)/(Lτ − − B), then in view of (3.19), (3.47), (3.58), and (3.59),
there exists a µ0 ∈ {1,2,... } such that for j = 1,2,...,m,
B mτ + δ
zr(κnµ0 +2φ+κ−s j )+τ ≤ C +
= mτ + δ
Lτ − (3.60)
and so from (3.19), (3.47), (3.48), (3.58), (3.59), and (3.60), we get
C Lτ − = Lτ − > l τ .
yr(κnµ0 +2φ+κ)+τ ≥ B +
mτ + δ
(3.61)
Using (3.19), (3.47), (3.48), (3.58), (3.59), and (3.61) and working inductively, we can
easily prove that
yr(κnµ0 +2φ+κω)+τ ≥ Lτ − > lτ ,
ω = 2,3,...,
(3.62)
which is a contradiction since liminf n→∞ yκrn+τ = lτ . Therefore, since τ is an arbitrary
number such that τ ∈ {0,1,...,κr − 1}, relations (3.45) are satisfied.
Moreover, from (3.19) and (3.47), we have that
lim zκnr+i = ξi ,
n→∞
i = 0,1,...,κr − 1.
(3.63)
This completes the proof of the lemma.
In the next proposition, we study the periodicity of the positive solutions of (1.1).
Proposition 3.7. Consider (1.1), where k,m ∈ {1,2,... }, A,ci ,d j , i ∈ {1,2,...,k}, j ∈
{1,2,...,m}, are positive fuzzy numbers, and pi , i ∈ {1,2,...,k }, q j , j ∈ {1,2,...,m}, are
positive integers. If (3.10) holds and r is a common divisor of the integers pi + 1, q j + 1,
i = 1,2,...,k, j = 1,2,...,m, then (1.1) has periodic solutions of prime period r. Moreover, if
ri , i = 1,2,...,k, (resp., s j , j = 1,2,...,m)—ri , s j are defined in (3.36)—are even (resp., odd)
integers, then (1.1) has periodic solutions of prime period 2r.
350
Fuzzy difference equations
Proof. From (3.10), we have that A, ci , i = 1,2,...,k, d j , j = 1,2,...,m, are positive real
numbers such that (3.12) and (3.13) hold. We consider functions Li,a ,Ri,a , i = −π, −π +
1,...,0, such that for λ = 0,1,...,φ − 1, θ = 1,2,...,r, and a ∈ (0,1],
L−rφ+rλ+θ,a = L−r+θ,a ,
R−rφ+rλ+θ,a = R−r+θ,a ,
(3.64)
the functions Lw,a , w = −r + 1, −r + 2,...,0, are increasing, left continuous, and for all
w = −r + 1, −r + 2,...,0, we have
A + < Lw,a < 2A,
Rw,a =
ALw,a
,
Lw,a − A
(3.65)
where is a positive number such that < A. Using (3.65) and since the functions Lw,a ,
w = −r + 1, −r + 2,...,0, are increasing, if a1 ,a2 ∈ (0,1], a1 ≤ a2 , we get
ALw,a1 Lw,a2 − A2 Lw,a1 ≥ ALw,a1 Lw,a2 − A2 Lw,a2
(3.66)
which implies that Rw,a , w = −r + 1, −r + 2,...,0, are decreasing functions. Moreover,
from (3.65), we get
Lw,a ≤ Rw,a ,
A + ≤ Lw,a ,Rw,a ≤
2A2
,
(3.67)
and so from [18, Theorem 2.1], Lw,a ,Rw,a , w = −r + 1, −r + 2,...,0, determine the fuzzy
numbers xw , w = −r + 1, −r + 2,...,0, such that [xw ]a = [Lw,a ,Rw,a ], w = −r + 1, −r +
2,...,0. Let xn be a positive solution of (1.1) which satisfies (2.14) and let the initial values
be positive fuzzy numbers such that (3.4) hold and the functions Li,a ,Ri,a , i = −π, −π +
1,...,0, a ∈ (0,1], are defined in (3.64), (3.65); Li,a , i = −π, −π + 1,...,0, a ∈ (0,1], are
increasing and left continuous. Then from [16, Proposition 2], we have that for any a ∈
(0,1], the system given by (3.7), (3.12), and
(3.13) has
periodic solutions of prime period
r, which means that there exists solution Ln,a ,Rn,a , a ∈ (0,1], of the system such that
Ln+r,a = Ln,a ,
Rn+r,a = Rn,a ,
a ∈ (0,1].
(3.68)
Therefore, from (2.22) and (3.68), we have that (1.1) has periodic solutions of prime
period r.
Now, suppose that ri , i = 1,2,...,k, (resp., si , j = 1,2,...,m) are even (resp., odd) integers. We consider the functions Li,a ,Ri,a , i = −π, −π + 1,...,0, such that analogous relations (3.37), (3.38), and (3.43) hold, Lw,a , w = −r + 1,...,0, are increasing, left continuous functions, and the first relation of (3.65) holds. Arguing as above, the solution
xn of (1.1) with initial values xi , i = −π, −π + 1,...,0, satisfying (3.4), where Li,a ,Ri,a ,
i = −π, −π + 1,...,0, are defined above, is a periodic solution of prime period 2r.
In the following proposition, we study the convergence of the positive solutions of
(1.1).
G. Stefanidou and G. Papaschinopoulos 351
Proposition 3.8. Consider (1.1), where k,m ∈ {1,2,... }, A,ci ,d j , i ∈ {1,2,...,k}, j ∈
{1,2,...,m}, are positive fuzzy numbers, and pi , i ∈ {1,2,...,k }, q j , j ∈ {1,2,...,m}, are
positive integers. Then the following statements are true.
(i) If (3.11), holds, then (1.1) has a unique positive equilibrium x and every positive
solution of (1.1) nearly converges to the unique positive equilibrium x with respect to D as
n → ∞ and converges to x with respect to D1 as n → ∞.
(ii) If (3.10) is satisfied and r is the greatest common divisor of the integers pi + 1, q j + 1,
i = 1,2,...,k, j = 1,2,...,m, such that (3.36) holds, then every positive solution of (1.1)
nearly converges to a period κr solution of (1.1) with respect to D as n → ∞ and converges to
a period κr solution of (1.1) with respect to D1 as n → ∞; κ is defined in (3.44).
Proof. (i) Let xn be a positive solution of (1.1) which satisfies (2.14). Since (3.7) and (3.11)
hold, we can apply [16, Proposition 4] and we have that for any a ∈ (0,1], there exist the
limn→∞ Ln,a , limn→∞ Rn,a , and
lim Ln,a = La ,
n→∞
lim Rn,a = Ra ,
n→∞
a ∈ (0,1],
(3.69)
where
La =
Al,a Ar,a − Ca Da
,
Ar,a − Ca
k
ci,l,a
,
d
j =1 j,r,a
Ca = mi=1
Ra =
k
Al,a Ar,a − Ca Da
,
Al,a − Da
ci,r,a
.
d
j =1 j,l,a
(3.70)
Da = mi=1
In addition, from (3.3) and (3.70), we get
La ≥
B2 − Z 2
= λ,
C−W
Ra ≤
C2 − W 2
= µ,
B−Z
(3.71)
where B,C (resp., Z,W) are defined in (3.3) (resp., (3.5)). Then from (3.69), (3.71), and
arguing as in [13, 14, 15], we can easily prove that La ,Ra determine a fuzzy number x such
that [x]a = [La ,Ra ]. Finally, using (3.70), we take that x is the unique positive equilibrium
of (1.1). Using relations (3.11), (3.69), and arguing as in [15, Proposition 2], we can prove
that every positive solution of (1.1) nearly converges to the unique positive equilibrium x
with respect to D as n → ∞ and converges to x with respect to D1 as n → ∞.
(ii) Suppose that (3.10) holds. Let xn be a positive solution of (1.1) such that (2.14)
holds. Since (Ln,a ,Rn,a ) is a positive solution of the system which is defined by (3.7),
(3.12), and (3.13), from Lemma 3.6, we have that
lim Lκnr+l,a = l,a ,
n→∞
lim Rκnr+l,a = ξl,a ,
n→∞
a ∈ (0,1], l = 0,1,...,κr − 1,
(3.72)
where κ is defined in (3.44). Using (3.72) and arguing as in [15, Proposition 2], we can
prove that every positive solution of (1.1) nearly converges to a period κr solution of (1.1)
with respect to D as n → ∞ and converges to a period κr solution of (1.1) with respect to
D1 as n → ∞. Thus, the proof of the proposition is completed.
From Propositions 3.2–3.8, it is obvious that (1.1) exhibits the trichotomy character
described concentratively by the following proposition.
352
Fuzzy difference equations
Proposition 3.9. Consider the fuzzy difference equation (1.1), where k,m ∈ {1,2,... }, and
A,ci ,d j , i ∈ {1,2,...,k}, j ∈ {1,2,...,m}, are positive fuzzy numbers. Then (1.1) possesses
the following trichotomy.
(i) If relation (3.1) is satisfied, then (1.1) has unbounded solutions.
(ii) If (3.10) holds and r is the greatest common divisor of the integers pi + 1, q j + 1,
i = 1,2,...,k, j = 1,2,...,m, such that (3.36) holds, then every positive solution of (1.1)
nearly converges to a period κr solution of (1.1) with respect to D as n → ∞ and converges to
a period κr solution of (1.1) with respect to D1 as n → ∞.
(iii) If (3.11) holds, then every positive solution of (1.1) nearly converges to the unique
positive equilibrium x with respect to D as n → ∞ and converges to x with respect to D1 as
n → ∞.
In the next proposition, we study the asymptotic stability of the unique positive equilibrium of (1.1).
Proposition 3.10. Consider the fuzzy difference equation (1.1), where k,m ∈ {1,2,... },
A,ci ,d j , i ∈ {1,2,...,k}, j ∈ {1,2,...,m}, are positive fuzzy numbers, and pi , i ∈ {1,2,...,k},
q j , j ∈ {1,2,...,m}, are positive integers such that (3.11) holds. Suppose that there exists a
positive number θ such that
θ < B,
Z<
2B + C − θ − (C − θ)2 + 4BC
,
2
(3.73)
where B,C are defined in (3.3) and Z is defined in (3.5). Then the unique positive equilibrium x of (1.1) is nearly asymptotically stable.
Proof. Since (3.11) holds, from Proposition 3.8, equation (1.1) has a unique positive
equilibrium x which satisfies (2.15).
Let be a positive real number. Since (3.18) holds, we can define the positive real
number δ as follows:
δ < min{,λ,θ,B − Z }.
(3.74)
Let xn be a positive solution of (1.1) such that
D(x−i ,x) ≤ δ ≤ ,
i = 0,1,...,π.
(3.75)
From (3.75), we have
L−i,a − La ≤ δ,
R−i,a − Ra ≤ δ,
i = 0,1,...,π, a ∈ (0,1].
(3.76)
In addition, from (3.3), (3.7), (3.74), and (3.76) and since (La ,Ra ) satisfies (3.7), we get
k
k
ci,l,a L− pi ,a
ci,l,a (La + δ)
− La ≤ Al,a + mi=1
− La
L1,a − La = Al,a + mi=1
j =1 d j,r,a R−q j ,a
j =1 d j,r,a (Ra − δ)
Ca − Al,a + La
Ra − (B − Z)
=δ
≤δ
.
Ra − δ
Ra − δ
(3.77)
G. Stefanidou and G. Papaschinopoulos 353
From (3.74) and (3.77), it is obvious that
L1,a − La < δ < .
(3.78)
Moreover, arguing as above, we can easily prove that
R1,a − Ra ≤ δ
Da − Ar,a + Ra
.
La − δ
(3.79)
We claim that
θ < La − Ra + Ar,a − Da ,
a ∈ (0,1].
(3.80)
We fix an a ∈ (0,1] and we concider the function
g(h) =
Al,a Ar,a − Da h Al,a Ar,a − Da h
−
+ Ar,a − Da ,
Ar,a − h
Al,a − Da
(3.81)
where h is a nonnegative real variable. Moreover, we consider the function
f (x, y,z) =
x2 − (2x + y)z + z2
− θ,
x−z
(3.82)
where B ≤ x ≤ y ≤ C and W ≤ z ≤ Z, B,C (resp., W,Z) are defined in (3.3) (resp., (3.5)).
Using (3.82), we can easily prove that the function f is increasing (resp., decreasing)
(resp., decreasing) with respect to x (resp., y) (resp., z) for all y,z (resp., x,z) (resp., x, y)
and so from (3.73),
f (x, y,z) > f (B,C,Z) =
B 2 − (2B + C)Z + Z 2
− θ > 0.
B−Z
(3.83)
Therefore, from (3.3), (3.81), (3.82), and (3.83), we have
g(0) = f Al,a ,Ar,a ,Da + θ > θ.
(3.84)
In addition, from (3.81), we can prove that g is an increasing function with respect to h
and so we have g(0) < g(Ca ), a ∈ (0,1]. Therefore, from (3.70), (3.81), and (3.84), relation
(3.80) is true. Hence, from (3.74), (3.79), and (3.80), we get
R1,a − Ra < δ < .
(3.85)
From (3.7), (3.76), (3.78), and (3.85) and working inductively, we can easily prove that
Ln,a − La ≤ ,
Rn,a − Ra ≤ ,
a ∈ (0,1], n = 0,1,...,
(3.86)
354
Fuzzy difference equations
and so
n ≥ 0.
D xn ,x ≤ ,
(3.87)
Therefore, the positive equilibrium x is stable. Moreover, from Proposition 3.8, we have
that every positive solution of (1.1) nearly tends to x with respect to D as n → ∞. So, x is
nearly asymptotically stable. So, the proof of the proposition is completed.
Finally, we study the oscillatory behavior of the positive solutions of the fuzzy difference equation
k
c2s+1 xn−2s−1
,
s=0 d2s+2 xn−2s
xn+1 = A + s=k 0
(3.88)
where k is a positive integer, and A, c2s+1 , d2s+2 , s ∈ {0,1,...,k}, are positive fuzzy numbers. Obviously, (3.88) is a special case of (1.1).
In what follows, we need to study the oscillatory behavior of the positive solutions of
the system of ordinary difference equations
k
a2s+1 yn−2s−1
,
s=0 b2s+2 zn−2s
yn+1 = B + s=k0
k
zn+1 = C +
s=0 e2s+1 zn−2s−1
,
k
s=0 h2s+2 yn−2s
n = 0,1,...,
(3.89)
where k is a positive integer, B,C,a2s+1 ,b2s+2 ,e2s+1 ,h2s+2 , s ∈ {0,1,...,k}, are positive real
constants, and the initial values y j , z j , j = −2k − 1, −2k,...,0, are positive real numbers.
Let (yn , zn ) be a positive solution of (3.89). We say that the solution (yn , zn ) oscillates
about (y, z), y,z ∈ R+ , if for every n0 ∈ N, there exist s, m ∈ N, s, m ≥ n0 , such that
ys − y ym − y ≤ 0,
ys − y zs − z ≥ 0,
zs − z zm − z ≤ 0,
ym − y zm − z ≥ 0.
(3.90)
Lemma 3.11. Consider system (3.89), where k is a positive integer, B, C, a2s+1 , b2s+2 , e2s+1 ,
h2s+2 , s ∈ {0,1,...,k}, are positive real constants, and the initial values y j , z j , j = −2k −
1, −2k,...,0, are positive real numbers.A positive
solution yn ,zn of system (3.89) oscillates
about the unique positive equilibrium x̄, ȳ of system (3.89) if either the relations
Λ ≥ max Λ1,s ,Λ2,s ,
∆ ≥ max ∆1,s ,∆2,s ,
s = 0,1,...,k,
(3.91)
s = 0,1,...,k,
(3.92)
or the relations
Λ ≤ min{Λ1,s ,Λ2,s ,
∆ ≤ min{∆1,s ,∆2,s ,
G. Stefanidou and G. Papaschinopoulos 355
hold, where for s = 0,1,...,k,
Λ=
∆1,s =
h2s+2
e2s+1
Λ2,s =
a2s+1 y−2s−1
,
s=0 b2s+2 z−2s
∆ = s=k0
1
1
s=0 e2s+1 z−2s−1
,
k
s=0 h2s+2 y−2s
s
k
s
−1
k
ȳ µ
b2 j+2 z̄+
b2 j+2 z−2 j+2+2s −
a2 j+1 ȳ+
a2 j+1 y−2 j+1+2s
z̄ j =0
j =s+1
j =0
j =s+1
a2s+1
∆2,s =
Λ1,s =
1
k
k
1
b2s+2
−1
k
s
−1
k
ȳ s
e2 j+1 z̄ + e2 j+1 z−2 j+2s −
h2 j+2 ȳ +
h2 j+2 y−2 j+1+2s
λz̄ j =0
j =s
j =0
j =s+1
s −1
k
s
−1
k
z̄ a2 j+1 ȳ+ a2 j+1 y−2 j+2s −
b2 j+2 z̄ +
b2 j+2 z−2 j+1+2s
µ ȳ j =0
j =s
j =0
j =s+1
k
− B,
− B,
s
k
s
−1
k
z̄ λ
h2 j+2 ȳ+
h2 j+2 y−2 j+2+2s −
e2 j+1 z̄+
e2 j+1 z−2 j+1+2s
ȳ j =0
j =s+1
j =0
j =s+1
− C,
− C,
k
e2s+1
,
s=0 h2s+2
a2s+1
.
s=0 b2s+2
λ = ks=0
µ = sk=0
(3.93)
Proof. Suppose that (3.91) hold. We prove that for ρ = 0,1,...,k,
y2ρ+1 ≥ ȳ,
z2ρ+1 ≥ z̄,
y2ρ+2 ≤ ȳ,
z2ρ+2 ≤ z̄.
(3.94)
From (3.89) and (3.91), we have
k
a2s+1 y−2s−1
= B + ∆ ≥ B + ∆1,k = ȳ,
s=0 b2s+2 z−2s
y1 = B + s=k0
(3.95)
z1 = C + Λ ≥ C + Λ1,k = z̄.
Since from (3.91), Λ ≥ Λ2,0 and ∆ ≥ ∆2,0 , then from (3.89), we have
k
y2 = B +
s=0 a2s+1 y−2s
b2 z1 + ks=1 b2s+2 z1−2s
µ ȳ
(C + Λ)b2 + ks=1 b2s+2 z1−2s µ ȳ
≤B+
=B+
= ȳ,
k
z̄
z̄
b2 z1 + s=1 b2s+2 z1−2s
z2 ≤ C +
λz̄
= z̄.
ȳ
(3.96)
Using (3.89), (3.91), (3.95), and (3.96), relations ∆ ≥ ∆1,ρ−1 , Λ ≥ Λ1,ρ−1 (resp., ∆ ≥
∆2,ρ , Λ ≥ Λ2,ρ ), ρ = 1,2,...,k, and working inductively, we can easily prove (3.94) for ρ =
1,2,...,k:
y2ρ+1 ≥ ȳ,
z2ρ+1 ≥ z̄
resp., y2ρ+2 ≤ ȳ, z2ρ+2 ≤ z̄ .
(3.97)
356
Fuzzy difference equations
Therefore, (3.94) hold for ρ = 0,1,...,k. Then since (3.94) hold for ρ = 0,1,...,k, using
(3.89) and working inductively, we can easily prove that(3.94) hold for any ρ = k + 1,k +
2,..., and so if (3.91) hold, the proof of the lemma is completed.
Similarly, if (3.92) are satisfied, then we can easily prove that
y2ρ+1 ≤ ȳ,
z2ρ+1 ≤ z̄,
y2ρ+2 ≥ ȳ,
z2ρ+2 ≥ z̄,
ρ = 0,1,....
(3.98)
This completes the proof of the lemma.
Using Lemma 3.11 and arguing as in [13, Proposition 2.4], we can easily prove the
following proposition which concerns the oscillatory behavior of the positive solutions of
the fuzzy difference equation (3.88).
Proposition 3.12. Consider (3.88), where k is a positive integer, and A, c2s+1 , d2s+2 , s ∈ {0,
1,...,k}, are positive fuzzy numbers. Then a positive solution xn of (3.88) satisfying (2.14)
oscillates about the positive equilibrium x, which satisfies (2.15) if, for any s = 0,1,...,k and
a ∈ (0,1], either the relations
Λ̄a ≥ max Λ̄1,s,a , Λ̄2,s,a ,
∆¯ a ≥ max ∆¯ 1,s,a , ∆¯ 2,s,a
(3.99)
or the relations
Λ̄a ≤ min Λ̄1,s,a , Λ̄2,s,a ,
∆¯ a ≤ min ∆¯ 1,s,a , ∆¯ 2,s,a
(3.100)
hold, where Λ̄a , ∆¯ a , Λ̄1,s,a , Λ̄2,s,a , ∆¯ 1,s,a , ∆¯ 2,s,a are defined for the analogous system (3.7) in the
same way as Λ, ∆, Λ1,s ,Λ2,s , ∆1,s ,∆2,s were defined in Lemma 3.11 for system (3.89).
Using Proposition 3.12, we take the following corollary.
Corollary 3.13. Consider (3.88), where k is a positive integer, and A, c2s+1 , d2s+2 , s ∈
{0,1,...,k }, are positive fuzzy numbers. Then a positive solution xn of (3.88) satisfying (2.14)
oscillates about the positive equilibrium x, which satisfies (2.15) if, for any p = 0,1,...,k and
a ∈ (0,1], either the relations
L−2k−1+2p,a ≥ La ,
L−2k+2p,a ≤ La ,
R−2k−1+2p,a ≥ Ra ,
R−2k+2p,a ≤ Ra
(3.101)
or the relations
L−2k−1+2p,a ≤ La ,
L−2k+2p,a ≥ La ,
R−2k−1+2p,a ≤ Ra ,
R−2k+2p,a ≥ Ra
hold.
Acknowledgment
This work is a part of the first author Doctoral thesis.
(3.102)
G. Stefanidou and G. Papaschinopoulos 357
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
R. P. Agarwal, Difference Equations and Inequalities, Monographs and Textbooks in Pure and
Applied Mathematics, vol. 155, Marcel Dekker, New York, 1992.
D. Benest and C. Froeschle, Analysis and Modelling of Discrete Dynamical Systems, Advances
in Discrete Mathematics and Applications, vol. 1, Gordon and Breach Science Publishers,
Amsterdam, 1998.
E. Y. Deeba and A. De Korvin, Analysis by fuzzy difference equations of a model of CO2 level in
the blood, Appl. Math. Lett. 12 (1999), no. 3, 33–40.
E. Y. Deeba, A. De Korvin, and E. L. Koh, A fuzzy difference equation with an application, J.
Differ. Equations Appl. 2 (1996), no. 4, 365–374.
p
q
R. DeVault, G. Ladas, and S. W. Schultz, On the recursive sequence xn+1 = A/xn + B/xn−1 , Advances in Difference Equations (Veszprém, 1995), Gordon and Breach, Amsterdam, 1997,
pp. 125–136.
, On the recursive sequence xn+1 = A/xn + 1/xn−2 , Proc. Amer. Math. Soc. 126 (1998),
no. 11, 3257–3261.
L. Edelstein-Keshet, Mathematical Models in Biology, The Random House/Birkhäuser Mathematics Series, Random House, New York, 1988.
S. N. Elaydi, An Introduction to Difference Equations, Undergraduate Texts in Mathematics,
Springer-Verlag, New York, 1996.
S. Heilpern, Fuzzy mappings and fixed point theorem, J. Math. Anal. Appl. 83 (1981), no. 2,
566–569.
G. J. Klir and B. Yuan, Fuzzy Sets and Fuzzy Logic: Theory and Applications, Prentice Hall PTR,
New Jersey, 1995.
V. L. Kocić and G. Ladas, Global Behavior of Nonlinear Difference Equations of Higher Order
with Applications, Mathematics and its Applications, vol. 256, Kluwer Academic Publishers
Group, Dordrecht, 1993.
H. T. Nguyen and E. A. Walker, A First Course in Fuzzy Logic, CRC Press, Florida, 1997.
G. Papaschinopoulos and B. K. Papadopoulos, On the fuzzy difference equation xn+1 = A + B/xn ,
Soft Comput. 6 (2002), 436–440.
, On the fuzzy difference equation xn+1 = A + xn /xn−m , Fuzzy Sets and Systems 129
(2002), no. 1, 73–81.
G. Papaschinopoulos and G. Stefanidou, Boundedness and asymptotic behavior of the solutions
of a fuzzy difference equation, Fuzzy Sets and Systems 140 (2003), no. 3, 523–539.
, Trichotomy of a system of two difference equations, J. Math. Anal. Appl. 289 (2004),
no. 1, 216–230.
M. L. Puri and D. A. Ralescu, Differentials of fuzzy functions, J. Math. Anal. Appl. 91 (1983),
no. 2, 552–558.
C. Wu and B. Zhang, Embedding problem of noncompact fuzzy number space E∼ (I), Fuzzy Sets
and Systems 105 (1999), no. 1, 165–169.
G. Stefanidou: Department of Electrical and Computer Engineering, Democritus University of
Thrace, 67100 Xanthi, Greece
E-mail address: tfele@yahoo.gr
G. Papaschinopoulos: Department of Electrical and Computer Engineering, Democritus University of Thrace, 67100 Xanthi, Greece
E-mail address: gpapas@ee.duth.gr